April 2024 Evaluation Changes: What’s New?

Written by CDCB , USDA Animal Genomics and Improvement Laboratory

March 07, 2024

April 2024 Evaluation Changes: What’s New?

The Council on Dairy Cattle Breeding (CDCB) and USDA Animal Genomics Improvement Laboratory (AGIL) announce enhancements in the U.S. dairy genetic evaluations on April 2, 2024.

New Variance Component Estimates and Variance Adjusted Weights for Health Traits

By Taylor Marie McWhorter, Kristen Parker Gaddis, Ezequiel Nicolazzi, and Paul VanRaden

Since the 2018 debut of CDCB evaluations for disease resistance, the number of health records in the National Cooperator Database has tripled or quadrupled – depending on the trait. With this data surge, the trait model has been adjusted with new variance component estimates and adjusted weights, effective with the April 2024 evaluations. This follows a typical progression and evolution of newer traits.  

CDCB genetic evaluations are provided for six direct health traits for the resistance to Milk Fever (MFEV), Displaced Abomasum (DA), Ketosis (KETO), Mastitis (MAST), Metritis (METR) and Retained Placenta (RETP). The health evaluations were first incorporated for Holstein in April 2018, Jersey in April 2020, and Brown Swiss in August 2022. Variance components were originally estimated in 2018, when Holstein records available for each trait ranged from 1.2 to 2.2 million. The available records by trait are summarized in the first table.

 Milk FeverDisplaced AbomasumKetosisMastitisMetritisRetained Placenta
# Records, Millions (HO) 2018 Estimate1.2 M1.9 M1.4 M2.4 M2.0 M2.2 M
# Records in Database, Millions (HO, JE, BS) Dec 20235.8 M5.8 M4.3 M7.7 M6.3 M7.6 M
Incidence Rates Dec 20231.06%1.63%5.84%11.72%7.16%3.39%

A single-trait linear animal repeatability model with random effects (additive, permanent environment, herd-by-year, and herd-by-sire) is used to estimate genomic Predicted Transmitting Ability (PTA) for animals. New variance components were estimated for all six health traits. The new estimates affect overall heritabilities (h2; all lactations together), h2-by-lactation, variance ratios for other random effects, and repeatability.

 Milk FeverDisplaced AbomasumKetosisMastitisMetritisRetained Placenta
h2  20180.6%1.1%1.2%3.1%1.4%1.0%
h2  20230.6%3.1%1.7%3.2%1.6%1.3%

In addition, weights applied to health traits were updated from 0 or 1 to be a value estimated from the variance components. These variance-adjusted weights are used to standardize genetic variance across differing parities that have differing heritabilities. The new weights are a function of repeatability, h2, and h2-by-lactation. The previous weights of 0 or 1 are now 0 or 0.25-1.46 depending on lactation and trait.

In evaluations, both the existing variance adjusted phenotypes and the new variance adjusted weights are employed to account for the heterogeneous variance.

To investigate the impact of the upgrades made to the health evaluations, a test using data from the December 2023 run was completed. Correlations between the official December 2023 evaluation and the test evaluation with the updates were examined for old and young animals by breed and for genomic estimated breeding values (GEBV) and genomic reliability (GREL). GEBV (GREL) correlations for MFEV, KETO, MAST, METR, and RETP were ≥0.96 (≥0.98) for Holstein, ≥0.90 (≥0.95) for Jersey and ≥0.92 (≥0.99) for Brown Swiss. The DA GEBV (GREL) correlations were ≥0.95 (≥0.93) For Holstein, ≥0.82 (≥0.84) for Jersey and ≥0.81 (≥0.96) for Brown Swiss. The lower correlations observed for DA are due to the largest change in h2, which also impacts the variance-adjusted weights. 

Given the new variance components and correlations between official and test run, some variation is expected in PTA for individual animals. However, the variance adjustments effectively capture the categorical trait of incidence in a linear model. The impact on Net Merit is expected to be very small, given the weighting of these traits in the index.

Foreign Unknown Parent Groups

By Paul VanRaden and Andres Legarra

Unknown Parent Groups (UPG) are used in the genetic evaluation to provide an average Predicted Transmitting Ability (PTA) value to all animals with missing pedigree information. Starting in the April 2024 evaluation, all foreign UPG effects will be merged with domestic UPGs.

Accurately estimating UPG effects requires that foreign unknown parents have descendants with domestic phenotypes. That has been the case with Canadian dams and international sires, which describe the first foreign-genotyped animals. As genotyping has expanded across the world, a high proportion of recently genotyped foreign unknown parents do not have descendants with domestic phenotypes. Thus, the automated system defining UPG has recently been detecting a much larger need for UPGs, alongside a drastic reduction in the available information to estimate them. Furthermore, a recent study showed the behavior of these foreign UPG could be introducing bias in the evaluation. For this reason, starting in the April 2024 evaluation, all foreign UPG effects will be merged with domestic UPGs. Foreign genetic trends were assumed equal to domestic trends, which is a reasonable assumption for Europe and Canada. Lack of information to estimate actual trends in South America or Asia, where most of the unknown parents are, makes this a need to reduce overall bias.

Before implementation, AGIL and CDCB tested the impacts and correlations. For example, fertility traits of Daughter Pregnancy Rate (DPR), Cow Conception Rate (CCR), Heifer Conception Rate (HCR), and Early First Calving (EFC) were compared with and without separate foreign UPGs in the pedigree model before adding genomic information. For Holstein bulls born in the last 10 years with >50% reliability, PTA correlations before and after merging the groups were >99.98% for all four traits. Across all years, PTA correlations were >99.97%, except EFC with correlation of 99.86% due to a 10% faster estimated genetic trend. For the EFC trait, the largest changes were for European bulls of breeds Simmental or Montbeliard, and for U.S. bulls with Canadian dam IDs but no further pedigree.

Generally, this change will have little effect on animal evaluations across the population, although there will be significant effect on specific animals that used the removed UPG.

Breed Base Representation (BBR) Update Delayed to August

By Ezequiel L. Nicolazzi

The Breed Base Representation (BBR) reference population update is typically updated on an annual basis, in the April evaluations. This year’s BBR update will be delayed for four months, as we are working to introduce a new SNP list. To avoid two consecutive changes to BBR reference animals, the new SNP set and update to the BBR reference population are scheduled to be implemented together in August 2024.

Interbull Validation on Fertility, Somatic Cell Score and Mastitis

By Rodrigo Mota and Taylor Marie McWhorter

Since 1995, the U.S. evaluation system has participated to the Multiple Across-Country Evaluation (MACE) for bulls, managed by Interbull. This multi-country evaluation is beneficial because it allows receipt of evaluations, on U.S. scale, for bulls with daughters in other countries. This enhances the comparison of performance of most of the bulls in the world, while improving the accuracy of parent averages of their U.S. progeny (if any). In order to exchange evaluations, countries are required to achieve Interbull validation that their evaluations are free from bias. This validation happens every two years, or when a trait has model changes, as is the case for Mastitis.*

CDCB fertility traits, Somatic Cell Score and Mastitis were due for a check-up in 2024, and all traits were successfully validated for all breeds.

*The model changes for Mastitis are described in the first section, “New variance component estimates and variance adjusted weights for health traits.”